Evidence and governance
How a number becomes publishable, and what stops it.
Who this is for
M&E officers, analysts and anyone approving a publication.
No screenshot on this pageThe evidence and lineage capture this guide carried was withdrawn from publication on 2026-09-22. Read by eye, it rendered a real administrator’s name and the real VoiceInsights Africa organisation rather than the demonstration tenant it had been published as. Re-capturing it on the demonstration tenant needs a staff sign-in, and Preview cannot currently deliver the email one-time code that sign-in requires, so no governed capture of this surface exists to publish. Putting a picture of an adjacent screen here instead would make a caption claim something the image does not show.
The evidence chain
A published figure is traceable back through claim → evidence → analysis run → dataset → study → instrument → responses. Provenance entries are hash-chained, so a broken or reordered chain is detectable rather than merely unlikely.
Promotion is written once
Promoting an analysis result into the evidence chain is idempotent: a repeat promotion of the same result into the same report writes nothing and says so. Lineage is a fact recorded once.
Retention, erasure and legal hold
These are an executable lifecycle — dry run, approval, governed execution, per-class audit record — not a settings page.
- A legal hold blocks an approved erasure, and holds are re-evaluated at execution time rather than at approval time.
- The absence of a retention policy means preserve. It never means permission to delete.
- Audit and evidence records cannot be given a deletion policy. The record of an erasure outlives the erasure.
What this does not do
Stated here rather than discovered later
- A finding whose evidence does not resolve does not publish. The export is refused rather than produced with a caveat.
- Promotion does not re-run the analysis. It records that a result already computed is now cited.
- Nothing here can be overridden. There is no bypass, and implying one exists would be worse than the objection it answers.